archomrade [he/him] @ archomrade @midwest.social

Posts

54

Comments

2,214

Joined

2 yr. ago

"They're all good, so good. Number 3: 'eat a frog'..... That sounds promising. Number 4: 'Eat a dog'.... I don't know, um, from a practical standpoint if that's--"

I thought about it, but I have a couple other services that could benefit from getting dedicated gpu anyway. Might as well just save for a proper PCIE card.

cool, i might try it

I wanted to get a dedicated card for video transcription anyway, but it's good to know I don't necessarily need it.

I'm curious what you're doing with frigate/ how you're doing it without a graphics card?

I've been using it for object detection, but i had to install it on my workhorse because my server doesn't have a graphics card. I suppose it doesn't need one if you're not doing ml processing, but I'm still curious

This might be controversial, but you and I both have the means to mass collect data, or find illicit datasets already collected. The kind of data collection that we don't have access to (the kind that's taken from your phone without your consent) isn't really helpful for training LLM's. But, again, if you have the means to replicate their methodology to begin with then you likely already have all of the material. You're not going to recreate their model on consumer hardware anyway.

They're just not advertising where that data is (and neither should anyone here)

if you have to break the law to be able to compile it yourself, its not foss.

Not if you consider apps like jellyfin or plex to be FOSS, but even that comparison is apples and oranges because training a model that big isn't something you can do on your own hardware. Just because they haven't given you the data to alter the model doesn't mean they haven't given you everything you need to use it with your own data and your own hardware. I get that people inherently distrust AI companies (and Chinese companies especially, but I won't get into that here), but I think it's misplaced here.

They aren't going to break the law for you. If you want to train your own LLM you'll have to source your own copyrighted dataset for the task.

Jellyfin doesn't come with a media library, you have to 'rip' your own dvd's and home videos. Same deal.

None of the flagship models publish their training data because they're all trained on less-than-legal datasets.

It's a little like complaining that jellyfin doesn't publish any media with their code - not only is that not legal but it's implied that you're responsible for attaining your own.

If you're someone who can and does compile and re-train your own 64B parameter LLM models, you almost certainly have your own dataset for that purpose (in fact huggingface has many).

Idk why people keep saying this - they published their methodology and the code that runs the model with the weights. The only things they didn't publish with it are likely copyrighted works that cant be freely shared. It's 'open-sourced' in all the ways that matter

And nvidia bounced after the US signaled intent to block or investigate deepseek, not necessarily because the model isn't a threat

Characterizing that as 'unawareness' is a little too charitable, I think.

Tldr - selfhosting is useful when:

• you need a lot of storage
• you need a lot of processing
• you are collaborating with multiple people/family members
• you are sharing media with other people outside your network
• you are sharing media across devices
• you want a standalone backup independent of your mobile device without doing so manually
• you want more advanced AI features that are not feasible to do on device (such as image detection or live security camera object detection)
• you want your home IOT devices to work locally without a cloud connection
• you have old hardware collecting dust and want to put it to use
• you like to make things

Seems like you might have understood the purpose of those apps, you just didn't personally have those needs yourself, and that's fine

Lol well this was fun but I have better things to do

The private password is on GitHub in plain text.

Right because that's a completely normal thing to do

Lmao, you'll have to do better than "experts see discrepancies in the data", because that's what Mike Lindell had, too.

Remote access code + the private admin password + the code to flip the votes

If this were even true, why would they put it on github, let alone with the password in plain text. Lol Jesus christ do you have any idea how ridiculous this theory is?

I don't much go for self-indulgent fantasies, thanks.

edit: i feel like this is like someone seeing a magic trick and immediately being convinced wizards are real.

Why would a clandestine foreign agent publish malware on a public code repository? Some random reddit user claims to have found a repo on github that uses a publicly known username tied to a politically embroiled tech company and now we're supposed to believe it was used to falsify an entire electoral system?

It doesn't even pass the sniff test bud, what credibility are we supposed to lend to these anonymous users?

Do you have any evidence that isn't based on the assumption that democratic voters simply wouldn't split their vote? Or the assumption that people wouldn't just vote for president and not any other offices?

Like, IP logs or recount discrepancies? Evidence of malware on the machines? Anything other than "this looks implausible"?

difference between the nonsense Trump pushed out and this

Trump and his allies cited exactly the same kind of 'anomalous voting trends' as evidence of vote manipulation. Unless you have something more substantive than 'these ballots don't look like we expected them to' then this is exactly the same kind of non-evidence MAGA had.

The biden administration was exceptionally unpopular. Anti-Kamala democratic voters have been very clear about why they didn't vote for her. Rather than reckoning with their complete unpopularity, democrats would rather blame their loss on 'woke' politics and vote manipulation.

After 4 years of liberals laughing away MAGA conspiracies about hacked voting machines in the 2020 election, suddenly those concerns are very serious and very real?

I haven't seen anything in the way of actual evidence something nefarious happened here, except some hefty speculation about split ticket voting and a couple vague (but entirely on-brand) comments from Trump.

It's just funny to me that liberals are unironically repeating the same baseless accusations that chuds were for the last 4 years without much more in the way of evidence (if any at all)

Evidence of Election Hacking

The lack of self-awareness here is absolutely astounding.

For whatever it's worth, I'm entertained by these bans and have no problem with your moderation style.

Users from the larger instances can be melodramatic - it's nice to see them get burned for it on occasion.

I'm just trying to follow your use of the word fanboy bud, take a chill pill

if approving a ban reversal for tiktok is 'fanboying' trump, then approving of biden for not being an out and about nazi seems like an equally obtuse of the word 'fanboying'

There's a hell of a lot of obfuscation happening in your word choices